Assessing the Generalizability of Randomized Evidence by Comparing Results From a Clinical Trial and Established Quality Improvement Collaborative: Results From G-MINOR and MUSIC

Document Type

Book Review

Publication Date


Publication Title

Journal of Urology


INTRODUCTION AND OBJECTIVE: Although randomized controlled trials (RCTs) represent the gold standard for clinical research, results from RCTs are often substantially less dramatic when implemented in practice. Previous studies have evaluated the congruence of observational studies with RCTs and shown conflicting results. However, minimal to no data exist comparing RCT populations with their non-RCT contemporaries. We aimed to assess the generalizability of RCT evidence by comparing results from an RCT with those from a prospectively collected, geographically equivalent, quality improvement collaborative . METHODS: The G-MINOR RCT randomized 356 men with localized prostate cancer at high-risk of recurrence after prostatectomy to clinical risk stratification with or without Decipher testing to assess its utility in clinical decision-making. Enrollment of patients occurred across12 practices within the Michigan Urological Surgery Improvement Collaborative (MUSIC). Matching between MUSIC and G-MINOR was performed using 3:1 nearest neighbor matching on propensity scores. Differences between the matched populations were assessed via Kaplan-Meier estimates and multivariable Cox regression, with a primary endpoint of failure-free survival (FFS), which was defined as any biochemical recurrence or receipt of salvage treatment. Secondary endpoints were adjuvant treatment-free survival and cumulative incidence of salvage treatment. RESULTS: After matching, a total of 1352 men were included in the analysis (338 G-MINOR, 1014 MUSIC), with no differences in age, race, comorbidity, PSA, grade group, clinical T-stage, margin status, presence of extra-prostatic extension, or Decipher risk between the groups. Despite matching, FFS was greater at 48 months follow-up in the G-MINOR group compared to MUSIC (20.1% vs 19.9%,p=0.041). Men in MUSIC received significantly more adjuvant therapy (20.1% vs 9.2%, p<0.0001), but salvage treatment rates were similar between the groups (G-MINOR 11.2% vs MUSIC 8.6%, p=0.62). CONCLUSIONS: Despite similar geographic, practice-level, and disease-related characteristics, men enrolled in a RCT were less likely to have failure after initial treatment compared to matched patients in the same communities not enrolled in the RCT. These results suggest possible unmeasured differences between these populations or unmeasured benefits of trial participation that may result in improved outcomes for RCT participants.





First Page



American Urological Association Annual Meeting, April 28 - May 1, 2023, Chicago, IL